You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The following, which passes on the CPU, fails on the GPU for reasons I do not understand. It corresponds to the tests marked as "broken" at time of posting on buildkite CI.
accel=CUDALibs() # passing for CPU1()
using MLJBase, MLJFlux
X, y = make_regression()
model = NeuralNetworkRegressor(accleration=accel)
model.epochs = 1
mach = machine(model, X, y);
# two epochs in stages:
Random.seed!(123)
fit!(mach, verbosity=0);
model.epochs = model.epochs + 1
fit!(mach, verbosity=0);
l1 = MLJBase.report(mach).training_losses[end]
# two epochs in one go:
Random.seed!(123)
fit!(mach, verbosity=1, force=true)
l2 = MLJBase.report(mach).training_losses[end]
@test l1 ≈ l2
The text was updated successfully, but these errors were encountered:
context: #86.
The following, which passes on the CPU, fails on the GPU for reasons I do not understand. It corresponds to the tests marked as "broken" at time of posting on buildkite CI.
The text was updated successfully, but these errors were encountered: